Stohastic_gradient_decent_Adam_iteration_time.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "•features - матрица, соответствующая обучающей выборке\n",
    "\n",
    "•w_init - вектор начальных весов модели\n",
    "\n",
    "•eta - шаг градиентного спуска (по умолчанию 0.001)\n",
    "\n",
    "•max_iter - максимальное число итераций градиентного спуска (по умолчанию 10000)\n",
    "\n",
    "•max_weight_dist - минимальное евклидово расстояние между векторами весов на соседних итерациях градиентного спуска, при котором алгоритм прекращает работу (по умолчанию 1e-8)\n",
    "\n",
    "•seed - число, используемое для воспроизводимости сгенерированных псевдослучайных чисел (по умолчанию 42)\n",
    "\n",
    "•verbose - флаг печати информации (например, для отладки, по умолчанию False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "from time import time\n",
    "\n",
    "from thbmodel.utils import timestamp, YEAR, MONTH,settings\n",
    "from thbmodel.utils.arrayview import ArrayView, TimeseriesView\n",
    "\n",
    "from thbmodel.prediction.models.fit_model import TSModel\n",
    "from thbmodel.prediction.models.model_parameters import ModelParameters\n",
    "from thbmodel.prediction.models.factor_management import FactorList\n",
    "from thbmodel.prediction.models.preprocessing import load_slices\n",
    "\n",
    "from thbmodel.prediction.models.prediction import factornames_trimmed\n",
    "from thbmodel.prediction.models.parameters import factor_build_end\n",
    "\n",
    "import thbmodel.prediction.models.St_gradient as sg "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "av = ArrayView.from_file(settings.paths.join('racing_data_azd.av.bcolz'))\n",
    "tsav = load_slices()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "build_end =float(timestamp('2016-04-01'))\n",
    "oos_start =float(timestamp('2017-05-01'))\n",
    "pars = ModelParameters(av, build_end =build_end, oos_start =oos_start, depth=1, lmbd=10, verbose=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "is1_start   2016-04-01 12:40:00.000000\n",
      "is1_end   2017-04-30 16:45:00.000000\n",
      "oos_start   2017-05-01 12:30:00.000000\n",
      "oos_end   2017-08-20 15:00:00.000000\n"
     ]
    }
   ],
   "source": [
    "fl = FactorList.from_av(av, factornames_trimmed)\n",
    "fl.preprocess(pars)\n",
    "factors = fl.asmatrix()\n",
    "print 'is1_start  ',(timestamp(av.start_time[pars.is1][0]))\n",
    "print 'is1_end  ',(timestamp(av.start_time[pars.is1][-1]))\n",
    "print 'oos_start  ',(timestamp(av.start_time[pars.oos][0]))\n",
    "print 'oos_end  ',(timestamp(av.start_time[pars.oos][-1]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>f1</th>\n",
       "      <th>f2</th>\n",
       "      <th>f3</th>\n",
       "      <th>f4</th>\n",
       "      <th>f5</th>\n",
       "      <th>f6</th>\n",
       "      <th>f7</th>\n",
       "      <th>f8</th>\n",
       "      <th>f9</th>\n",
       "      <th>f10</th>\n",
       "      <th>f11</th>\n",
       "      <th>f12</th>\n",
       "      <th>f13</th>\n",
       "      <th>f14</th>\n",
       "      <th>f15</th>\n",
       "      <th>f16</th>\n",
       "      <th>f17</th>\n",
       "      <th>f18</th>\n",
       "      <th>f19</th>\n",
       "      <th>f20</th>\n",
       "      <th>f21</th>\n",
       "      <th>f22</th>\n",
       "      <th>f23</th>\n",
       "      <th>f24</th>\n",
       "      <th>f25</th>\n",
       "      <th>f26</th>\n",
       "      <th>f27</th>\n",
       "      <th>f28</th>\n",
       "      <th>f29</th>\n",
       "      <th>f30</th>\n",
       "      <th>f31</th>\n",
       "      <th>f32</th>\n",
       "      <th>f33</th>\n",
       "      <th>f34</th>\n",
       "      <th>f35</th>\n",
       "      <th>f36</th>\n",
       "      <th>f37</th>\n",
       "      <th>f38</th>\n",
       "      <th>f39</th>\n",
       "      <th>f40</th>\n",
       "      <th>f41</th>\n",
       "      <th>f42</th>\n",
       "      <th>f43</th>\n",
       "      <th>f44</th>\n",
       "      <th>f45</th>\n",
       "      <th>f46</th>\n",
       "      <th>f47</th>\n",
       "      <th>f48</th>\n",
       "      <th>f49</th>\n",
       "      <th>f50</th>\n",
       "      <th>f51</th>\n",
       "      <th>f52</th>\n",
       "      <th>f53</th>\n",
       "      <th>f54</th>\n",
       "      <th>event_id</th>\n",
       "      <th>result</th>\n",
       "      <th>is1</th>\n",
       "      <th>is2</th>\n",
       "      <th>oos</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.367198</td>\n",
       "      <td>-2.822814</td>\n",
       "      <td>1.042788</td>\n",
       "      <td>-1.449000</td>\n",
       "      <td>-0.399689</td>\n",
       "      <td>1.519560</td>\n",
       "      <td>1.120796</td>\n",
       "      <td>-0.903438</td>\n",
       "      <td>2.149808</td>\n",
       "      <td>0.454737</td>\n",
       "      <td>0.397418</td>\n",
       "      <td>2.018940</td>\n",
       "      <td>-0.041120</td>\n",
       "      <td>-1.864951</td>\n",
       "      <td>0.903440</td>\n",
       "      <td>-0.595091</td>\n",
       "      <td>-2.351100</td>\n",
       "      <td>0.791641</td>\n",
       "      <td>-0.371619</td>\n",
       "      <td>-0.134965</td>\n",
       "      <td>0.383271</td>\n",
       "      <td>-1.827844</td>\n",
       "      <td>-0.662651</td>\n",
       "      <td>1.423997</td>\n",
       "      <td>0.518412</td>\n",
       "      <td>0.633018</td>\n",
       "      <td>-0.217998</td>\n",
       "      <td>-0.046744</td>\n",
       "      <td>1.832861</td>\n",
       "      <td>0.362147</td>\n",
       "      <td>1.789398</td>\n",
       "      <td>0.713323</td>\n",
       "      <td>0.847637</td>\n",
       "      <td>2.197710</td>\n",
       "      <td>0.333612</td>\n",
       "      <td>0.992382</td>\n",
       "      <td>2.005222</td>\n",
       "      <td>1.745112</td>\n",
       "      <td>-2.169849</td>\n",
       "      <td>0.024419</td>\n",
       "      <td>0.516602</td>\n",
       "      <td>0.838355</td>\n",
       "      <td>-0.231825</td>\n",
       "      <td>0.510499</td>\n",
       "      <td>-0.818574</td>\n",
       "      <td>1.690850</td>\n",
       "      <td>1.153778</td>\n",
       "      <td>0.0</td>\n",
       "      <td>-0.178424</td>\n",
       "      <td>1.606775</td>\n",
       "      <td>1.276101</td>\n",
       "      <td>0.903439</td>\n",
       "      <td>-0.903432</td>\n",
       "      <td>1.539463</td>\n",
       "      <td>307976</td>\n",
       "      <td>4</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.367198</td>\n",
       "      <td>-2.822814</td>\n",
       "      <td>1.146508</td>\n",
       "      <td>-0.950907</td>\n",
       "      <td>0.134657</td>\n",
       "      <td>1.279091</td>\n",
       "      <td>0.709838</td>\n",
       "      <td>0.511180</td>\n",
       "      <td>1.163663</td>\n",
       "      <td>0.554670</td>\n",
       "      <td>0.589283</td>\n",
       "      <td>2.676868</td>\n",
       "      <td>-0.853920</td>\n",
       "      <td>-1.762160</td>\n",
       "      <td>-0.511180</td>\n",
       "      <td>0.355150</td>\n",
       "      <td>0.325790</td>\n",
       "      <td>-0.024739</td>\n",
       "      <td>0.462871</td>\n",
       "      <td>0.297558</td>\n",
       "      <td>1.503070</td>\n",
       "      <td>-1.925740</td>\n",
       "      <td>-0.642401</td>\n",
       "      <td>2.296713</td>\n",
       "      <td>-0.518494</td>\n",
       "      <td>1.678214</td>\n",
       "      <td>-0.325527</td>\n",
       "      <td>0.081803</td>\n",
       "      <td>-0.721549</td>\n",
       "      <td>3.691803</td>\n",
       "      <td>1.126171</td>\n",
       "      <td>1.576948</td>\n",
       "      <td>0.329637</td>\n",
       "      <td>0.138260</td>\n",
       "      <td>1.866920</td>\n",
       "      <td>1.105528</td>\n",
       "      <td>1.685527</td>\n",
       "      <td>1.745112</td>\n",
       "      <td>-1.926777</td>\n",
       "      <td>0.024419</td>\n",
       "      <td>-0.399449</td>\n",
       "      <td>0.838355</td>\n",
       "      <td>1.043211</td>\n",
       "      <td>0.510499</td>\n",
       "      <td>-0.881552</td>\n",
       "      <td>0.215314</td>\n",
       "      <td>0.760771</td>\n",
       "      <td>0.0</td>\n",
       "      <td>-0.103089</td>\n",
       "      <td>1.421764</td>\n",
       "      <td>0.382060</td>\n",
       "      <td>-0.511180</td>\n",
       "      <td>0.511188</td>\n",
       "      <td>-0.341251</td>\n",
       "      <td>307976</td>\n",
       "      <td>1</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.367198</td>\n",
       "      <td>-2.269686</td>\n",
       "      <td>1.109573</td>\n",
       "      <td>1.041469</td>\n",
       "      <td>-1.246235</td>\n",
       "      <td>0.959507</td>\n",
       "      <td>0.533713</td>\n",
       "      <td>-0.834292</td>\n",
       "      <td>2.054531</td>\n",
       "      <td>0.589805</td>\n",
       "      <td>0.781147</td>\n",
       "      <td>-0.425354</td>\n",
       "      <td>-0.308430</td>\n",
       "      <td>-1.495752</td>\n",
       "      <td>0.834292</td>\n",
       "      <td>-0.561211</td>\n",
       "      <td>0.570014</td>\n",
       "      <td>-1.385371</td>\n",
       "      <td>0.230456</td>\n",
       "      <td>0.295614</td>\n",
       "      <td>0.221700</td>\n",
       "      <td>-1.316387</td>\n",
       "      <td>-1.368199</td>\n",
       "      <td>0.486579</td>\n",
       "      <td>1.954420</td>\n",
       "      <td>0.373406</td>\n",
       "      <td>-0.280811</td>\n",
       "      <td>0.210350</td>\n",
       "      <td>0.824291</td>\n",
       "      <td>0.099820</td>\n",
       "      <td>0.924145</td>\n",
       "      <td>0.695130</td>\n",
       "      <td>0.329637</td>\n",
       "      <td>-2.094361</td>\n",
       "      <td>0.463911</td>\n",
       "      <td>1.114437</td>\n",
       "      <td>1.611524</td>\n",
       "      <td>1.745112</td>\n",
       "      <td>-1.271665</td>\n",
       "      <td>-0.100011</td>\n",
       "      <td>-0.118486</td>\n",
       "      <td>-1.065489</td>\n",
       "      <td>-0.231825</td>\n",
       "      <td>0.510499</td>\n",
       "      <td>2.544475</td>\n",
       "      <td>-0.108367</td>\n",
       "      <td>-0.777575</td>\n",
       "      <td>0.0</td>\n",
       "      <td>-0.034824</td>\n",
       "      <td>0.366740</td>\n",
       "      <td>0.442908</td>\n",
       "      <td>0.834292</td>\n",
       "      <td>-0.834286</td>\n",
       "      <td>0.687687</td>\n",
       "      <td>307976</td>\n",
       "      <td>2</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.367198</td>\n",
       "      <td>-2.082636</td>\n",
       "      <td>-0.546706</td>\n",
       "      <td>1.041469</td>\n",
       "      <td>-0.177928</td>\n",
       "      <td>0.807879</td>\n",
       "      <td>1.825297</td>\n",
       "      <td>0.377624</td>\n",
       "      <td>-0.891473</td>\n",
       "      <td>2.359050</td>\n",
       "      <td>0.690858</td>\n",
       "      <td>-1.025506</td>\n",
       "      <td>2.586447</td>\n",
       "      <td>-1.542717</td>\n",
       "      <td>-0.377623</td>\n",
       "      <td>0.483454</td>\n",
       "      <td>0.439499</td>\n",
       "      <td>-1.929624</td>\n",
       "      <td>-0.011781</td>\n",
       "      <td>0.218581</td>\n",
       "      <td>1.028220</td>\n",
       "      <td>-0.193052</td>\n",
       "      <td>-0.390140</td>\n",
       "      <td>-0.525911</td>\n",
       "      <td>0.543346</td>\n",
       "      <td>0.061765</td>\n",
       "      <td>-0.134161</td>\n",
       "      <td>0.210350</td>\n",
       "      <td>-1.481083</td>\n",
       "      <td>-1.201241</td>\n",
       "      <td>-0.343101</td>\n",
       "      <td>-0.052652</td>\n",
       "      <td>-0.188364</td>\n",
       "      <td>0.024540</td>\n",
       "      <td>-0.598031</td>\n",
       "      <td>-0.847758</td>\n",
       "      <td>1.648525</td>\n",
       "      <td>1.745112</td>\n",
       "      <td>-0.854125</td>\n",
       "      <td>-0.057719</td>\n",
       "      <td>-0.967518</td>\n",
       "      <td>-1.950998</td>\n",
       "      <td>-0.231825</td>\n",
       "      <td>0.510499</td>\n",
       "      <td>-0.705213</td>\n",
       "      <td>-0.063721</td>\n",
       "      <td>0.946241</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.028652</td>\n",
       "      <td>0.979326</td>\n",
       "      <td>3.985849</td>\n",
       "      <td>-0.377623</td>\n",
       "      <td>0.377629</td>\n",
       "      <td>1.750346</td>\n",
       "      <td>307976</td>\n",
       "      <td>5</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.367198</td>\n",
       "      <td>-1.620791</td>\n",
       "      <td>0.904502</td>\n",
       "      <td>1.041469</td>\n",
       "      <td>-0.645139</td>\n",
       "      <td>0.037615</td>\n",
       "      <td>-0.229496</td>\n",
       "      <td>-0.471971</td>\n",
       "      <td>1.330716</td>\n",
       "      <td>1.717160</td>\n",
       "      <td>0.167181</td>\n",
       "      <td>-0.232777</td>\n",
       "      <td>-0.308430</td>\n",
       "      <td>-0.513871</td>\n",
       "      <td>0.471972</td>\n",
       "      <td>-0.296116</td>\n",
       "      <td>0.195127</td>\n",
       "      <td>-1.385371</td>\n",
       "      <td>0.358433</td>\n",
       "      <td>0.311468</td>\n",
       "      <td>1.782793</td>\n",
       "      <td>-1.105682</td>\n",
       "      <td>-0.162607</td>\n",
       "      <td>-0.525911</td>\n",
       "      <td>1.667236</td>\n",
       "      <td>1.301977</td>\n",
       "      <td>-0.517036</td>\n",
       "      <td>0.210350</td>\n",
       "      <td>0.340718</td>\n",
       "      <td>4.356093</td>\n",
       "      <td>0.522335</td>\n",
       "      <td>1.517120</td>\n",
       "      <td>-0.188364</td>\n",
       "      <td>-1.785856</td>\n",
       "      <td>1.200810</td>\n",
       "      <td>0.854792</td>\n",
       "      <td>0.270584</td>\n",
       "      <td>1.745112</td>\n",
       "      <td>-0.640691</td>\n",
       "      <td>0.049978</td>\n",
       "      <td>-0.749274</td>\n",
       "      <td>-0.453683</td>\n",
       "      <td>-0.231825</td>\n",
       "      <td>0.510499</td>\n",
       "      <td>2.587987</td>\n",
       "      <td>-0.043631</td>\n",
       "      <td>1.434019</td>\n",
       "      <td>0.0</td>\n",
       "      <td>-0.145211</td>\n",
       "      <td>1.003482</td>\n",
       "      <td>0.699005</td>\n",
       "      <td>0.471972</td>\n",
       "      <td>-0.471966</td>\n",
       "      <td>0.365180</td>\n",
       "      <td>307976</td>\n",
       "      <td>8</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "         f1        f2        f3        f4        f5        f6        f7  \\\n",
       "0  0.367198 -2.822814  1.042788 -1.449000 -0.399689  1.519560  1.120796   \n",
       "1  0.367198 -2.822814  1.146508 -0.950907  0.134657  1.279091  0.709838   \n",
       "2  0.367198 -2.269686  1.109573  1.041469 -1.246235  0.959507  0.533713   \n",
       "3  0.367198 -2.082636 -0.546706  1.041469 -0.177928  0.807879  1.825297   \n",
       "4  0.367198 -1.620791  0.904502  1.041469 -0.645139  0.037615 -0.229496   \n",
       "\n",
       "         f8        f9       f10       f11       f12       f13       f14  \\\n",
       "0 -0.903438  2.149808  0.454737  0.397418  2.018940 -0.041120 -1.864951   \n",
       "1  0.511180  1.163663  0.554670  0.589283  2.676868 -0.853920 -1.762160   \n",
       "2 -0.834292  2.054531  0.589805  0.781147 -0.425354 -0.308430 -1.495752   \n",
       "3  0.377624 -0.891473  2.359050  0.690858 -1.025506  2.586447 -1.542717   \n",
       "4 -0.471971  1.330716  1.717160  0.167181 -0.232777 -0.308430 -0.513871   \n",
       "\n",
       "        f15       f16       f17       f18       f19       f20       f21  \\\n",
       "0  0.903440 -0.595091 -2.351100  0.791641 -0.371619 -0.134965  0.383271   \n",
       "1 -0.511180  0.355150  0.325790 -0.024739  0.462871  0.297558  1.503070   \n",
       "2  0.834292 -0.561211  0.570014 -1.385371  0.230456  0.295614  0.221700   \n",
       "3 -0.377623  0.483454  0.439499 -1.929624 -0.011781  0.218581  1.028220   \n",
       "4  0.471972 -0.296116  0.195127 -1.385371  0.358433  0.311468  1.782793   \n",
       "\n",
       "        f22       f23       f24       f25       f26       f27       f28  \\\n",
       "0 -1.827844 -0.662651  1.423997  0.518412  0.633018 -0.217998 -0.046744   \n",
       "1 -1.925740 -0.642401  2.296713 -0.518494  1.678214 -0.325527  0.081803   \n",
       "2 -1.316387 -1.368199  0.486579  1.954420  0.373406 -0.280811  0.210350   \n",
       "3 -0.193052 -0.390140 -0.525911  0.543346  0.061765 -0.134161  0.210350   \n",
       "4 -1.105682 -0.162607 -0.525911  1.667236  1.301977 -0.517036  0.210350   \n",
       "\n",
       "        f29       f30       f31       f32       f33       f34       f35  \\\n",
       "0  1.832861  0.362147  1.789398  0.713323  0.847637  2.197710  0.333612   \n",
       "1 -0.721549  3.691803  1.126171  1.576948  0.329637  0.138260  1.866920   \n",
       "2  0.824291  0.099820  0.924145  0.695130  0.329637 -2.094361  0.463911   \n",
       "3 -1.481083 -1.201241 -0.343101 -0.052652 -0.188364  0.024540 -0.598031   \n",
       "4  0.340718  4.356093  0.522335  1.517120 -0.188364 -1.785856  1.200810   \n",
       "\n",
       "        f36       f37       f38       f39       f40       f41       f42  \\\n",
       "0  0.992382  2.005222  1.745112 -2.169849  0.024419  0.516602  0.838355   \n",
       "1  1.105528  1.685527  1.745112 -1.926777  0.024419 -0.399449  0.838355   \n",
       "2  1.114437  1.611524  1.745112 -1.271665 -0.100011 -0.118486 -1.065489   \n",
       "3 -0.847758  1.648525  1.745112 -0.854125 -0.057719 -0.967518 -1.950998   \n",
       "4  0.854792  0.270584  1.745112 -0.640691  0.049978 -0.749274 -0.453683   \n",
       "\n",
       "        f43       f44       f45       f46       f47  f48       f49       f50  \\\n",
       "0 -0.231825  0.510499 -0.818574  1.690850  1.153778  0.0 -0.178424  1.606775   \n",
       "1  1.043211  0.510499 -0.881552  0.215314  0.760771  0.0 -0.103089  1.421764   \n",
       "2 -0.231825  0.510499  2.544475 -0.108367 -0.777575  0.0 -0.034824  0.366740   \n",
       "3 -0.231825  0.510499 -0.705213 -0.063721  0.946241  0.0  0.028652  0.979326   \n",
       "4 -0.231825  0.510499  2.587987 -0.043631  1.434019  0.0 -0.145211  1.003482   \n",
       "\n",
       "        f51       f52       f53       f54  event_id  result   is1   is2    oos  \n",
       "0  1.276101  0.903439 -0.903432  1.539463    307976       4  True  True  False  \n",
       "1  0.382060 -0.511180  0.511188 -0.341251    307976       1  True  True  False  \n",
       "2  0.442908  0.834292 -0.834286  0.687687    307976       2  True  True  False  \n",
       "3  3.985849 -0.377623  0.377629  1.750346    307976       5  True  True  False  \n",
       "4  0.699005  0.471972 -0.471966  0.365180    307976       8  True  True  False  "
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import pandas as pd\n",
    "pd.set_option('display.max_columns', 90)\n",
    "n,m = factors.shape\n",
    "col_names = ['f{}'.format(i) for i in range(1,n+1)]\n",
    "\n",
    "df = sg.DF (pars.model_mask, factors, av, col_names, ['event_id', 'result'])\n",
    "df['is1'] = pars.is1 [pars.model_mask]\n",
    "df['is2'] = pars.is2 [pars.model_mask]\n",
    "df['oos'] = pars.oos [pars.model_mask]\n",
    "\n",
    "\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "DataFrame to tensors 27.9991397858\n"
     ]
    }
   ],
   "source": [
    "import time\n",
    "start = time.time()\n",
    "df_train_t, winner_tr_t, t_len_tr = sg.DF_to_tensor(df.loc[df.is1.values,:], col_names)\n",
    "df_test_t, winner_ts_t, t_len_ts = sg.DF_to_tensor(df.loc[df.oos.values,:], col_names)\n",
    "end = time.time()\n",
    "print 'DataFrame to tensors {}'.format(end -start)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 98,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def AdaMax_iter_time(data, features,  w = None, b_size =100, eta=1e-3, \n",
    "                                moment1 = 0.85, moment2 = 0.99, la = 10., N_epoh =10, direction =False,\n",
    "                                eps = 1e-4,  min_weight_dist=1e-3, seed=42, verbose=False):\n",
    "    \"\"\" \n",
    "    градиентный спуск \n",
    "    при каждой итерации проходим все скачки от новых к старым c размером b_size один раз\n",
    "    переходим на следущую итерацию и повторяем проход по скачкам\n",
    "    \n",
    "    data <= табдица с данными содержащая len(features)-столбцов данных, столбец с номерами - event_id, \n",
    "            столбец с результатом - result\n",
    "    features <=  наименование признаков в данных \n",
    "    w <=  начальное приближение для параметров модели\n",
    "    b_size  <=  размер бача\n",
    "    eta <=  шаг обучения\n",
    "    moment1 <=  первый момент\n",
    "    moment2 <=  второй момент\n",
    "    la <=  коффициент регуляризации\n",
    "    max_iter <=  максимальное число итераций\n",
    "    direction <= True if forward False if back\n",
    "    eps <=  почти ноль  \n",
    "    min_weight_dist <=  признак сходимости параметров \n",
    "    seed <=  рандомная константа \n",
    "    verbose <=  флаг отладки\n",
    "    \"\"\"\n",
    "    \n",
    "    # Инициализируем расстояние между векторами весов на соседних\n",
    "    # итерациях большим числом. \n",
    "    weight_dist = np.inf\n",
    "    # Будем порождать псевдослучайные числа \n",
    "    # (номер объекта, который будет менять веса), а для воспроизводимости\n",
    "    # этой последовательности псевдослучайных чисел используем seed.\n",
    "    #np.random.seed(seed)\n",
    "    if w == None:\n",
    "         w = np.random.normal(0, scale =1e-2, size =(len(features)))\n",
    "    \n",
    "    # Step\n",
    "    iter_num = 0\n",
    "    \n",
    "    # last gradient value\n",
    "    last_grad = np.zeros_like(w) \n",
    "    # last gradient's variation value\n",
    "    std_grad = np.zeros_like(w)\n",
    "    # unique events in data set\n",
    "    events_list = np.unique(data['event_id'])\n",
    "    step_max = len (events_list)/b_size\n",
    "    # store of grad changes\n",
    "    grad_norm = np.zeros(N_epoh *step_max +1)\n",
    "    # Main loop\n",
    "    for  i_ep  in range(1, N_epoh+1):\n",
    "                \n",
    "        for step_ in range(1, step_max +1):\n",
    "            \n",
    "            if direction:\n",
    "                r_event = events_list[b_size*(step_): b_size*(step_+1)]\n",
    "            else:\n",
    "                r_event = events_list[-b_size*(step_ +1): -b_size*step_]\n",
    "            df = data.loc[np.in1d(data['event_id'], r_event),:]\n",
    "            \n",
    "            df_t, winner_t, t_len = DF_to_tensor(df, features)\n",
    "            \n",
    "            w_N = w - eta *moment1 *last_grad # for Nesterov momentum\n",
    "            # get gradient with  Nesterov momemtum + L2 regularisation\n",
    "            grad_ = G_step(df_t, winner_t, t_len, w_N, verbose=False) + la * w_N  # gradient and L2\n",
    "            #grad_ = G_step(df_t, winner_t, t_len, w, verbose=False) + la * w  # gradient and L2\n",
    "            last_grad = moment1 * last_grad + (1-moment1) * grad_ # Update weight first moment estimate\n",
    "            # Update weight second moment estimate exponent power infinity\n",
    "            std_grad = np.max(np.vstack((moment2 * std_grad, np.abs(grad_))), axis =0) # Update weight second moment estimate \n",
    "            \n",
    "            t = (i_ep -1)*step_max + step_\n",
    "            last_grad_ = last_grad/(1 - np.power(moment1, t)) # Correct first moment estimate\n",
    "        \n",
    "            w = w - eta *last_grad_/std_grad\n",
    "        \n",
    "            weight_dist = np.linalg.norm(last_grad, ord=2)/len(w)\n",
    "            grad_norm [t] = weight_dist\n",
    "            \n",
    "            if (weight_dist <= min_weight_dist):\n",
    "                print 'SGD covergence', t\n",
    "                break\n",
    "                return w, grad_norm , iter_num*step_\n",
    "        \n",
    "            if verbose:\n",
    "                #print 'w  ', w\n",
    "                print 'iteration', iter_num, 'dist ',weight_dist\n",
    "         \n",
    "            if np.any((w == np.inf ) |(w == -np.inf)):\n",
    "                print \"w overcome\"\n",
    "                break\n",
    "        \n",
    "    return w, grad_norm , t"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def AdaMax_time_step(data, features,  w = None, b_size =100, eta=1e-3, \n",
    "                                moment1 = 0.85, moment2 = 0.99, la = 10., max_iter =10,\n",
    "                                eps = 1e-4,  min_weight_dist=1e-3, seed=42, verbose=False):\n",
    "    \"\"\" \n",
    "    градиентный спуск \n",
    "    берем b_size скачек более новых и делаем с ними max_iter итераций \n",
    "    затем следующие b_size скачек старее и делаем с ними max_iter итераций\n",
    "    пока на переберем все скачки\n",
    "    \n",
    "    data <= табдица с данными содержащая len(features)-столбцов данных, столбец с номерами - event_id, \n",
    "            столбец с результатом - result\n",
    "    features <=  наименование признаков в данных \n",
    "    w <=  начальное приближение для параметров модели\n",
    "    b_size  <=  размер бача\n",
    "    eta <=  шаг обучения\n",
    "    moment1 <=  первый момент\n",
    "    moment2 <=  второй момент\n",
    "    la <=  коффициент регуляризации\n",
    "    max_iter <=  максимальное число итераций\n",
    "    eps <=  почти ноль  \n",
    "    min_weight_dist <=  признак сходимости параметров \n",
    "    seed <=  рандомная константа \n",
    "    verbose <=  флаг отладки\n",
    "    \"\"\"\n",
    "\n",
    "    \n",
    "    # Инициализируем расстояние между векторами весов на соседних\n",
    "    # итерациях большим числом. \n",
    "    weight_dist = np.inf\n",
    "    # для воспроизводимости\n",
    "    # этой последовательности псевдослучайных чисел используем seed.\n",
    "    #np.random.seed(seed)\n",
    "    if w == None:\n",
    "        w = np.random.normal(0, scale =1e-2, size =(len(features)))\n",
    "    \n",
    "    # Step\n",
    "    step_ = 0\n",
    "    \n",
    "    # last gradient value\n",
    "    last_grad = np.zeros_like(w) \n",
    "    # last gradient's variation value\n",
    "    std_grad = np.zeros_like(w)\n",
    "    # unique events in data set\n",
    "    events_list = np.unique(data['event_id'])\n",
    "    step_max = len (events_list)/b_size\n",
    "    # store of grad changes\n",
    "    grad_norm = np.zeros(max_iter *step_max +1)\n",
    "    # Main loop\n",
    "    for step_ in range (1, step_max +1):\n",
    "        \n",
    "        r_event = events_list[-b_size*(step_ +1): -b_size*step_]\n",
    "        df = data.loc[np.in1d(data['event_id'], r_event),:]\n",
    "        \n",
    "        df_t, winner_t, t_len = DF_to_tensor(df, features)\n",
    "        for iter_num  in range(1, max_iter +1):\n",
    "            \n",
    "            w_N = w - eta *moment1 *last_grad # for Nesterov momentum\n",
    "            # get gradient with  Nesterov momemtum + L2 regularisation\n",
    "            grad_ = G_step(df_t, winner_t, t_len, w_N, verbose=False) + la * w_N  # gradient and L2 \n",
    "            #grad_ = G_step(df_t, winner_t, t_len, w, verbose=False) + la * w  # gradient and L2 \n",
    "            last_grad = moment1 * last_grad + (1-moment1) * grad_ # Update weight first moment estimate\n",
    "            # Update weight second moment estimate expanent power infinity\n",
    "            std_grad = np.max(np.vstack((moment2 * std_grad, np.abs(grad_))), axis =0) # Update weight second moment estimate \n",
    "        \n",
    "            t = max_iter *(step_-1) + iter_num\n",
    "            last_grad_ = last_grad/(1 - np.power(moment1, t)) # Correct first moment estimate\n",
    "        \n",
    "            w = w - eta *last_grad_/std_grad\n",
    "        \n",
    "            weight_dist = np.linalg.norm(last_grad, ord=2)/len(w)\n",
    "            grad_norm [t]   = weight_dist\n",
    "            \n",
    "        if (weight_dist <= min_weight_dist):\n",
    "            print 'SGD covergence', t\n",
    "            break\n",
    "            return w, grad_norm , iter_num*step_\n",
    "        \n",
    "        if verbose:\n",
    "            #print 'w  ', w\n",
    "            print 'iteration', iter_num, 'dist ',weight_dist\n",
    "        \n",
    "        if np.any(w == np.nan):\n",
    "            print \"w overcome\"\n",
    "            break\n",
    "        \n",
    "    return w, grad_norm , t"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "steps   138\n"
     ]
    }
   ],
   "source": [
    "events_list = np.unique(df.loc[df.is1.values,'event_id'])\n",
    "print 'steps  ', len(events_list)/100"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<module 'thbmodel.prediction.models.St_gradient' from 'thbmodel/prediction/models/St_gradient.py'>"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import imp\n",
    "imp.reload(sg)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "SGD covergence 1\n",
      "Time taken to stohastic gradient 184.514117002\n"
     ]
    }
   ],
   "source": [
    "import time\n",
    "start = time.time()\n",
    "\n",
    "SGD3_f = sg.AdaMax_iter_time(df.loc[df.is1.values,:], col_names, b_size =100, eta = 1e-3, moment1 = 0.97, \\\n",
    "                                         moment2 =0.99,  la = 0.8, N_epoh =10, direction =True, verbose=False)\n",
    "\n",
    "end = time.time()\n",
    "print 'Time taken to stohastic gradient {}'.format(end - start)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "SGD covergence 1\n",
      "Time taken to stohastic gradient 179.944425106\n"
     ]
    }
   ],
   "source": [
    "import time\n",
    "start = time.time()\n",
    "\n",
    "SGD3_b = sg.AdaMax_iter_time(df.loc[df.is1.values,:], col_names, b_size =100, eta = 1e-3, moment1 = 0.97, \\\n",
    "                                         moment2 =0.99,  la = 0.8, N_epoh =10, direction =False, verbose=False)\n",
    "\n",
    "end = time.time()\n",
    "print 'Time taken to stohastic gradient {}'.format(end - start)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "steps   13\n"
     ]
    }
   ],
   "source": [
    "events_list = np.unique(df.loc[df.is1.values,'event_id'])\n",
    "print 'steps  ', len(events_list)/1000"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "SGD covergence 300\n",
      "Time taken to stohastic gradient 6.88228106499\n"
     ]
    }
   ],
   "source": [
    "import time\n",
    "start = time.time()\n",
    "\n",
    "SGD4 = sg.AdaMax_time_step(df.loc[df.is1.values,:], col_names, b_size =1000, eta = 1e-3, moment1 = 0.97, \\\n",
    "                                         moment2 =0.99,  la = 0.8, max_iter =100, verbose=False)\n",
    "\n",
    "end = time.time()\n",
    "print 'Time taken to stohastic gradient {}'.format(end - start)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAABJEAAAJCCAYAAABwNFYJAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fd4917d9c10>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "ename": "NameError",
     "evalue": "name 'SGD3' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-29-273adbdc6601>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m      5\u001b[0m \u001b[0mplt\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlegend\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      6\u001b[0m \u001b[0mplt\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mshow\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 7\u001b[0;31m \u001b[0;32mprint\u001b[0m \u001b[0mSGD3_b\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mSGD3\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m>\u001b[0m \u001b[0;36m0\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m-\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      8\u001b[0m \u001b[0;32mprint\u001b[0m \u001b[0mSGD3_f\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mSGD3\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m>\u001b[0m \u001b[0;36m0\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m-\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      9\u001b[0m \u001b[0;32mprint\u001b[0m \u001b[0mSGD4\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mSGD4\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m>\u001b[0m \u001b[0;36m0\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m-\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mNameError\u001b[0m: name 'SGD3' is not defined"
     ]
    }
   ],
   "source": [
    "plt.figure(figsize=(20, 10))\n",
    "plt.plot(SGD3_b[1], label ='iereation_time_back')\n",
    "plt.plot(SGD3_f[1], label ='iereation_time_forward')\n",
    "plt.plot(SGD4[1], label ='time_step')\n",
    "plt.legend()\n",
    "plt.show()\n",
    "print SGD3_b[1][SGD3[1]> 0][-1]\n",
    "print SGD3_f[1][SGD3[1]> 0][-1]\n",
    "print SGD4[1][SGD4[1]> 0][-1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "w0 = SGD3_f[0] # return to original shape\n",
    "start = time.time()\n",
    "likelihood = sg.LL_t (df_train_t, winner_tr_t, w0)\n",
    "end = time.time()\n",
    "print 'count ll for train data {} by Adam_iter_time'.format(end -start)\n",
    "print likelihood"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "w0 = SGD3_b[0] # return to original shape\n",
    "start = time.time()\n",
    "likelihood = sg.LL_t (df_train_t, winner_tr_t, w0)\n",
    "end = time.time()\n",
    "print 'count ll for train data {} by Adam_iter_time'.format(end -start)\n",
    "print likelihood"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 84,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "count ll for train data 0.0498261451721 by Adam_time_step\n",
      "-1.99495426658\n"
     ]
    }
   ],
   "source": [
    "w0 = SGD4[0] # return to original shape\n",
    "start = time.time()\n",
    "likelihood = sg.LL_t (df_train_t, winner_tr_t, w0)\n",
    "end = time.time()\n",
    "print 'count ll for train data {} by Adam_time_step'.format(end -start)\n",
    "print likelihood"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 85,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "count ll for test data 0.0139899253845 by Adam_iter_time\n",
      "-1.99849806695\n"
     ]
    }
   ],
   "source": [
    "w0 = SGD3_f[0] # return to original shape\n",
    "start = time.time()\n",
    "likelihood = sg.LL_t (df_test_t, winner_ts_t, w0)\n",
    "end = time.time()\n",
    "print 'count ll for test data {} by Adam_iter_time'.format(end -start)\n",
    "print likelihood"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "count ll for test data 0.0290970802307 by Adam_iter_time\n",
      "-2.00551363843\n"
     ]
    }
   ],
   "source": [
    "w0 = SGD3_b[0] # return to original shape\n",
    "start = time.time()\n",
    "likelihood = sg.LL_t (df_test_t, winner_ts_t, w0)\n",
    "end = time.time()\n",
    "print 'count ll for test data {} by Adam_iter_time'.format(end -start)\n",
    "print likelihood"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 87,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "count ll for test data 0.0274538993835 by Adam_time_step\n",
      "-2.00409685347\n"
     ]
    }
   ],
   "source": [
    "w0 = SGD4[0] # return to original shape\n",
    "start = time.time()\n",
    "likelihood = sg.LL_t (df_test_t, winner_ts_t, w0)\n",
    "end = time.time()\n",
    "print 'count ll for test data {} by Adam_time_step'.format(end -start)\n",
    "print likelihood"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train   -2.02427213499\n",
      "test   -2.02850240438\n",
      "train   -2.01594138785\n",
      "test   -2.02236116846\n",
      "train   -2.02133836841\n",
      "test   -2.02840198534\n",
      "train   -2.04109180742\n",
      "test   -2.04337431746\n",
      "train   -2.02385636721\n",
      "test   -2.03031722461\n",
      "train   -2.0261073981\n",
      "test   -2.03530076029\n",
      "train   -2.03223450265\n",
      "test   -2.03733933133\n",
      "train   -2.02455212414\n",
      "test   -2.03231205185\n",
      "train   -2.01761160293\n",
      "test   -2.02539962818\n",
      "train   -2.0218015718\n",
      "test   -2.02765410657\n"
     ]
    }
   ],
   "source": [
    "for i in range(10):\n",
    "    \n",
    "    W1 = np.random.normal(loc =w0, scale =1e-2, size =(len(col_names)))\n",
    "    print 'train  ',LL_f (df_train_t, winner_tr_t, W1)\n",
    "    print 'test  ',LL_f (df_test_t, winner_ts_t, W1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "anaconda-cloud": {},
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}